Intensive intervention delivery in autism and early prediction of cognitive outcomes in children using computerised game and wireless EEG in naturalistic setting

Abstract

Neuro-developmental disorders, like Autism Spectrum Disorder, are common in youths and their incidence is increasing over the last decades. Although these disorders are attributed to disrupted brain functioning, there is no agreed biological marker for their diagnosis and treatment, thus they are based on behavioural analysis. Combating autism relies in three particular strands: early diagnosis, intensive personalised intervention (25 hours/week), and effective monitoring strategy of the children to tailor the therapy depending upon evolution of their developmental trajectories.

In this dissertation we contribute to each of these aspects as follows: We develop a multi-player game platform (GOLIAH) that enables delivering intensive intervention, not only in clinical, but also at-home settings. It implements the most recent ESDM intervention protocol targeting two "pivotal" skills that promote the learning process. In addition, it automatically generates behavioural measures when the child goes through the intervention process (playing the game). The operational procedure of GOLIAH was validated in a 3-month open trial conducted at Pitie-Salpetriere Hospital and Stella Maris Foundation. We develop a combined developmental and neuro-biological marker strategy for monitoring the trajectory of the children during the intervention.

The behavioural parameters for monitoring are based on the performance parameters of the child while playing GOLIAH. For the neurological markers we used wireless EEG-derived functional connectivity networks and its associated parameters. However, since the EEG recorded in mobile environment is corrupted by artefacts, as a first step for EEG processing, we develop two novel artefact reduction algorithms, WPTEMD and WPTICA. Through detailed analysis of semi-simulated and real EEG data and comparison with state-of-the-art algorithms, we establish that WPTEMD works best for wireless EEG data in naturalistic settings.

We explore the possibility of using functional brain connectivity networks for early detection of cognitive disabilities in an at-risk population (neonates with HIE). Our analysis shows that two years cognitive outcomes could be predicted from EEG recorded within two weeks of life with 87.5% of accuracy. We expect that over the first few months of life, a periodic monitoring of the functional brain connectivity networks and its parameters could help in improving the final outcome prediction.

More information

Identifiers

Catalogue record

Export record